書き換えによる抄録の読みやすさの向上  [in Japanese] Producing More Readable Extracts by Revising Them  [in Japanese]

    • 難波 英嗣 NANBA Hidetsugu
    • 北陸先端科学技術大学院大学情報科学研究科 School of Information Science, Japan Advanced Institute of Science and Technology
    • 奥村 学 OKUMURA Manabu
    • 北陸先端科学技術大学院大学情報科学研究科 School of Information Science, Japan Advanced Institute of Science and Technology

Abstract

これまで,抄録を自動的に作成する様々な手法が提案されてきた[9].その多くは,テキスト中の文を1つの単位とし,それらに何らかの情報を付与し,その重要度で順序付け,重要な文を選択し,それらを寄せ集めることで抄録を作成する.しかし,こうして作成された抄録の問題点として,Paice[6]は, 1)抽出した文中に代名詞などが含まれている場合,その先行詞が要約文中に存在する保証がないこと, 2)テキスト中の色々な箇所から抽出したものを単に集めているため,抽出した複数の文間のつながり(首尾一貫性)が悪いことなどを指摘している. このような読みにくさを解消する1つの方法に抄録の書き換えがある.抄録中の欠落情報を補ったり不要な箇所を削除するといった書き換えを行うことで文間のつながりが良くなり,抄録の読みにくさが解消されると考えられる.本研究では,まず心理実験で数人の被験者に抄録を読みやすく書き換えてもらい,抄録の読みにくさの要因を5種類に分析した.また読みにくさを解消する方法を検討した.次に分析結果に基づき抄録を読みやすく書き換えるシステムを計算機上に実装した.

Several techniques have been proposed for summarizing a text[9]. Each sentence in a text is considered as a unit for extraction, and summarization system calculates the importance of each sentence using several extract methods. Then several sentences are extracted according to their importance. However, Paice[6] points out the problem of the coherence of extracts as follows: 1) If anaphors are included in the extracts, their antecedents are not always included in the extracts. 2) Extracts are incoherent, because they are constructed from several sentences (phrases) which are partially extracted from texts. In order to make the extracts more readable, first, we have to clear the reasons of incoherence of the extracts. We let some subjects revise the extracts. Then, we classify the reasons of incoherence into 5 categories. Furthermore, we make rules for revising the extracts, and evaluate them by comparing the extracts revised by rules with those by subjects. Then, we find that our rules are effective to revise readability of the extracts

Journal

IPSJ SIG Notes   [List of Volumes]

IPSJ SIG Notes 99(73), 53-60, 1999-09-10  [Table of Contents]

Information Processing Society of Japan (IPSJ)

References:  15

You must have a user ID to see the references.If you already have a user ID, please click "Login" to access the info.New users can click "Sign Up" to register for an user ID.

Cited by:  11

You must have a user ID to see the cited references.If you already have a user ID, please click "Login" to access the info.New users can click "Sign Up" to register for an user ID.

Preview

Preview

Codes

  • NII Article ID (NAID) :
    110002935160
  • NII NACSIS-CAT ID (NCID) :
    AN10115061
  • Text Lang :
    JPN
  • Article Type :
    Journal Article
  • ISSN :
    09196072
  • NDL Article ID :
    5340224
  • NDL Source Classification :
    ZM13(科学技術--科学技術一般--データ処理・計算機)
  • NDL Call No. :
    Z14-1121
  • Databases :
    CJP  CJPref  NDL  NII-ELS 

Share